The Data Analytics Process

The data analytics process is the sequence of steps that turns a business question into a data-backed answer. Different books use slightly different names, but the standard university sequence is these seven steps.

Note the feedback arrows — the process is iterative, not a one-way waterfall.

Step 1 — Define the Problem / Objective

Everything downstream is wasted if this step is vague. Convert a fuzzy business ask into a precise, measurable analytical question.

Deliverables of this step: objective statement, success metric, scope, constraints, and the stakeholders who will act on the result.

Step 2 — Data Collection

Identify and gather the data needed to answer the question — from databases, APIs, files, surveys, web scraping, sensors, or third-party providers. (Covered in detail in the Data Collection lesson.)

Step 3 — Data Cleaning and Preparation

Typically 60–80% of total project time. Handle missing values, duplicates, outliers, inconsistent formats and wrong data types; then integrate, reduce and transform the data. (Covered in the Data Cleaning, Data Preprocessing and Data Transformation lessons.)

Step 4 — Data Exploration (EDA)

Summarise and visualize the prepared data to understand distributions, relationships and anomalies before modelling. (Unit 2.)

Step 5 — Data Modelling and Analysis

Apply the appropriate statistical or machine learning technique — regression, classification, clustering, association rules, hypothesis tests. (Units 2 and 3.)

Step 6 — Interpretation and Visualization

Translate model output into plain-language insight, with charts and dashboards aimed at the decision-maker, not at the analyst.

Step 7 — Deployment and Decision Making

Put the insight to work: a report, a dashboard, a policy change, or a model deployed into production — then monitor whether it actually improved the metric defined in Step 1.

A Miniature End-to-End Example

import pandas as pd

# STEP 1 — Question: "Which weekday has the weakest canteen sales, and by how much?"

# STEP 2 — Collection
data = {
    "day":   ["Mon", "Tue", "Wed", "Thu", "Fri", "Wed", "Mon"],
    "sales": [1200, 1450, 980, 1600, 2100, None, 1250],
}
df = pd.DataFrame(data)

# STEP 3 — Cleaning: one missing sales value
df["sales"] = df["sales"].fillna(df["sales"].median())

# STEP 4 — Exploration
print(df.groupby("day")["sales"].mean().sort_values())
# Wed     1202.5
# Mon     1225.0
# Tue     1450.0
# Thu     1600.0
# Fri     2100.0

# STEP 5 — Analysis
avg = df["sales"].mean()
worst = df.groupby("day")["sales"].mean().idxmin()
gap = avg - df.groupby("day")["sales"].mean().min()
print(f"Weakest day: {worst}, below overall average by Rs.{gap:.2f}")
# Weakest day: Wed, below overall average by Rs.259.64

# STEP 6 — Visualization
# df.groupby("day")["sales"].mean().plot(kind="bar")

# STEP 7 — Decision: run a Wednesday combo offer, then re-measure next month.

Common Pitfalls at Each Step

Process vs Lifecycle

Students often confuse the data analytics process with the data analytics lifecycle. The process is the general workflow of analysis; the lifecycle (next lesson) is a formal project-management framework with defined phases, entry/exit criteria and team roles. The lifecycle wraps around the process.